MatchScore

packages/providers/ContactsProvider/src/com/android/providers/contacts/aggregation/util/MatchScore.java

这个类用于记录每个联系人匹配分数，自动聚合的时候依据这个来选取候选对象。

public class MatchScore implements Comparable<MatchScore> {// Scores a multiplied by this number to allow room for "fractional" scorespublic static final int SCORE_SCALE = 1000; //分数所占系数// Best possible match scorepublic static final int MAX_SCORE = 100; //最大分数private long mRawContactId; //联系人相关信息private long mContactId;private long mAccountId;private boolean mKeepIn; //是否匹配private boolean mKeepOut;private int mPrimaryScore; //首要分数，名字匹配分数private int mSecondaryScore; //次要分数，号码等其它信息匹配分数private int mMatchCount;  //每次更新分数后值加1，最终也会用于计算匹配分数...
}

接下俩是显示成员变量关系的方法：

    public int getScore() {if (mKeepOut) {return 0; //不匹配直接返回0}if (mKeepIn) {return MAX_SCORE; //匹配直接返回100}int score = (mPrimaryScore > mSecondaryScore ? mPrimaryScore : mSecondaryScore); //选取最大的score// Ensure that of two contacts with the same match score the one with more matching// data elements wins.return score * SCORE_SCALE + mMatchCount; //系数是1000.可见mMatchCount占的比例很小}

比较方法

    @Overridepublic int compareTo(MatchScore another) {return another.getScore() - getScore(); //比较分数值}

NameDistance

packages/providers/ContactsProvider/src/com/android/providers/contacts/aggregation/util/NameDistance.java

这个类的方法就一个

public float getDistance(byte bytes1[], byte bytes2[]) 

返回两个名字的距离，注意参数都是byte格式的。这个距离的定义见 匹配算法Jaro–Winkler distance简介

ContactMatcher

packages/providers/ContactsProvider/src/com/android/providers/contacts/aggregation/util/ContactMatcher.java

联系人匹配分数计算

常量

    // Suggest to aggregate contacts if their match score is equal or greater than this thresholdpublic static final int SCORE_THRESHOLD_SUGGEST = 50; // Automatically aggregate contacts if their match score is equal or greater than this thresholdpublic static final int SCORE_THRESHOLD_PRIMARY = 70;// Automatically aggregate contacts if the match score is equal or greater than this threshold// and there is a secondary match (phone number, email etc).public static final int SCORE_THRESHOLD_SECONDARY = 50; 

三个常量，确定联系人匹配程度的阀值，值越低匹配的程度越低。还有其它的常量：

    private static final int NO_DATA_SCORE = -1; //不匹配的分数private static final int PHONE_MATCH_SCORE = 71; //号码匹配分数private static final int EMAIL_MATCH_SCORE = 71; //邮件匹配分数private static final int NICKNAME_MATCH_SCORE = 71; //昵称匹配分数private static final int MAX_MATCHED_NAME_LENGTH = 30; //最大匹配联系人数量

匹配算法常量，匹配程度由高到低：

    public static final int MATCHING_ALGORITHM_EXACT = 0; //完全匹配public static final int MATCHING_ALGORITHM_CONSERVATIVE = 1; //保守匹配public static final int MATCHING_ALGORITHM_APPROXIMATE = 2; //相近匹配

最后是

    public static final float APPROXIMATE_MATCH_THRESHOLD = 0.82f; //名字距离阀值public static final float APPROXIMATE_MATCH_THRESHOLD_FOR_EMAIL = 0.95f; //邮件地址距离阀值

成员

    private static int[] sMinScore =new int[NameLookupType.TYPE_COUNT * NameLookupType.TYPE_COUNT]; //最低分private static int[] sMaxScore =new int[NameLookupType.TYPE_COUNT * NameLookupType.TYPE_COUNT]; //最高分

在静态块中初始化：

    static {setScoreRange(NameLookupType.NAME_EXACT,NameLookupType.NAME_EXACT, 99, 99);setScoreRange(NameLookupType.NAME_VARIANT,NameLookupType.NAME_VARIANT, 90, 90);setScoreRange(NameLookupType.NAME_COLLATION_KEY,NameLookupType.NAME_COLLATION_KEY, 50, 80);...setScoreRange(NameLookupType.NICKNAME,NameLookupType.NICKNAME, 50, 60);setScoreRange(NameLookupType.NICKNAME,NameLookupType.NAME_COLLATION_KEY, 50, 60);setScoreRange(NameLookupType.NICKNAME,NameLookupType.EMAIL_BASED_NICKNAME, 50, 60);}

setCcoreRange方法，依据两中类型计算索引，并赋值，将一维数组当二维用:

    private static void setScoreRange(int candidateNameType, int nameType, int scoreFrom, int scoreTo) {int index = nameType * NameLookupType.TYPE_COUNT + candidateNameType;sMinScore[index] = scoreFrom;sMaxScore[index] = scoreTo;}

看MatchScore的集合成员：

    private final HashMap<Long, MatchScore> mScores = new HashMap<Long, MatchScore>();  //依据contact id获取MatchScoreprivate final ArrayList<MatchScore> mScoreList = new ArrayList<MatchScore>(); //MatchScore列表private int mScoreCount = 0; //匹配的分数个数，可能小于mScoreList的size

getMatchingScore方法展现了它们的关系

    private MatchScore getMatchingScore(long contactId) {MatchScore matchingScore = mScores.get(contactId); //先从缓存中获取if (matchingScore == null) {  //没有的话开始创建新的MatchScoreif (mScoreList.size() > mScoreCount) {  列表数目大于mScoreCount，则取一个元素并初始化matchingScore = mScoreList.get(mScoreCount);matchingScore.reset(contactId);} else {matchingScore = new MatchScore(contactId); //创建新的对象mScoreList.add(matchingScore);}mScoreCount++; //创建一个新的MatchScore后数量加1mScores.put(contactId, matchingScore); //放置对应contact id的MatchScore}return matchingScore;}

方法

  public void updateScoreWithPhoneNumberMatch(long contactId) {updateSecondaryScore(contactId, PHONE_MATCH_SCORE);}public void updateScoreWithEmailMatch(long contactId) {updateSecondaryScore(contactId, EMAIL_MATCH_SCORE);}public void updateScoreWithNicknameMatch(long contactId) {updateSecondaryScore(contactId, NICKNAME_MATCH_SCORE);}private void updatePrimaryScore(long contactId, int score) {getMatchingScore(contactId).updatePrimaryScore(score);}private void updateSecondaryScore(long contactId, int score) {getMatchingScore(contactId).updateSecondaryScore(score);}public void keepIn(long contactId) {getMatchingScore(contactId).keepIn();}public void keepOut(long contactId) {getMatchingScore(contactId).keepOut();}

一系列的更新MatchScore的方法

 public void matchName(long contactId, int candidateNameType, String candidateName,int nameType, String name, int algorithm) {int maxScore = getMaxScore(candidateNameType, nameType); //分数矩阵转换分数是0分，无需继续匹配if (maxScore == 0) {return;}if (candidateName.equals(name)) { //名字完全匹配updatePrimaryScore(contactId, maxScore);return;}if (algorithm == MATCHING_ALGORITHM_EXACT) { //算法是完全匹配，无需继续进行return;}int minScore = getMinScore(candidateNameType, nameType);if (minScore == maxScore) {//最小分和最大分相等，无需进行return;}final byte[] decodedCandidateName;final byte[] decodedName;try {decodedCandidateName = Hex.decodeHex(candidateName); //转换成byte，以供后续计算decodedName = Hex.decodeHex(name);} catch (RuntimeException e) {// How could this happen??  See bug 6827136Log.e(TAG, "Failed to decode normalized name.  Skipping.", e);return;}NameDistance nameDistance = algorithm == MATCHING_ALGORITHM_CONSERVATIVE ?mNameDistanceConservative : mNameDistanceApproximate;int score;float distance = nameDistance.getDistance(decodedCandidateName, decodedName); //计算名字举例boolean emailBased = candidateNameType == NameLookupType.EMAIL_BASED_NICKNAME|| nameType == NameLookupType.EMAIL_BASED_NICKNAME;float threshold = emailBased? APPROXIMATE_MATCH_THRESHOLD_FOR_EMAIL: APPROXIMATE_MATCH_THRESHOLD;if (distance > threshold) {score = (int)(minScore +  (maxScore - minScore) * (1.0f - distance)); //计算分数} else {score = 0;}updatePrimaryScore(contactId, score); //更新主要分数}

matchName是计算名字匹配分数方法

public List<Long> prepareSecondaryMatchCandidates(int threshold)

返回符合次要分数符合要求的联系人id列表

public List<MatchScore> pickBestMatches(int threshold) 

返回符合要求的MatchScore列表

public long pickBestMatch(int threshold, boolean allowMultipleMatches) 

返回最匹配的联系人id，就是分数最高的那个，如果有相同的最高分，则依据allowMultipleMatches返回不同的值

RawContactMatcher

packages/providers/ContactsProvider/src/com/android/providers/contacts/aggregation/util/RawContactMatcher.java

和ContactMatcher基本一样，最大的改动是方法参数中基本都加了account id。这个类只有ContactAggregator2类在使用，ContactMatcher是ContactAggregator使用。新版本的聚合是用ContactAggregator2。ContactsProvider2中依据算法版本号值不同创建不同的对象：

private void initForDefaultLocale() {...PROPERTY_AGGREGATION_ALGORITHM_VERSION = (value == 0)? AGGREGATION_ALGORITHM_OLD_VERSION: AGGREGATION_ALGORITHM_NEW_VERSION;mContactAggregator = (value == 0)? new ContactAggregator(this, mContactsHelper,createPhotoPriorityResolver(context), mNameSplitter, mCommonNicknameCache): new ContactAggregator2(this, mContactsHelper,createPhotoPriorityResolver(context), mNameSplitter, mCommonNicknameCache);...
}

目前是新版本号，也就是使用ContactAggregator2。

RawContactMatchingCandidates

packages/providers/ContactsProvider/src/com/android/providers/contacts/aggregation/util/RawContactMatchingCandidates.java

该类就是保存匹配联系人的一些相关数据，ContactAggregator2中用到

    private List<MatchScore> mBestMatches; //保存MatchScoreprivate Set<Long> mRawContactIds = null; //全部contact idprivate Map<Long, Long> mRawContactToContact = null; //RawContact id对应的Contact idprivate Map<Long, Long> mRawContactToAccount = null; //RawContact id对应的Account id

CommonNicknameCache

packages/providers/ContactsProvider/src/com/android/providers/contacts/aggregation/util/CommonNicknameCache.java
使用NICKNAME_LOOKUP表的封装，见联系人存储ContactsProvider表分析中的NICKNAME_LOOKUP表，例如"Robert", "Bob"and "Rob"这三个属于同一个CLUSTER，表示相同的名字，只不过写法不同（西方语言是表音文字，同一读音有多个写法，中文就无此问题了）。注意这个和data表中mime类型为Nickname的数据是两回事。